Assessing empirical linear and non-linear modelling of residual chlorine in urban drinking water systems

Abstract

To evaluate the performance of chlorine disinfection in drinking water plants, managers use as an indicator the concentrations of residual chlorine within the distribution system. Because of the residence time of water in storage tanks and in distribution pipelines, data that reflect residual chlorine concentrations for a given applied dose are only available after a certain time delay. Consequently, applied chlorine doses that are either too high or too low are often identified too late for an operator's reaction (either decreasing or increasing the applied doses) to be effective. To improve effectiveness in this area, modelling of residual chlorine appears to be an interesting alternative. This paper presents the application of two empirical models for simulating and forecasting residual chlorine concentrations within urban water systems. The first is a linear autoregressive model with external inputs, known as ARX; the second is a non-linear artificial neural network (ANN) model. The development of both models is founded on representative data from two Canadian drinking water systems. The results demonstrate the potential of an ANN model, which has a unique ability to detect non-linear complex relationships between data. In evaluating all the given data, simulation results show a similar performance for the linear and non-linear models. However, for specific water treatment conditions (very high and very low chlorine doses), the ANN model gives better predictions than the ARX model. Strategies designed to identify more representative data for future research are also proposed.